Amazon experiment

[Demesnedenoir]

So, the "Products Related to this Item" is based entirely on advertising, or so I had heard. What does this mean? It means that the book with the highest Product Display Ad bid appears first in the row.

So, one of my test ads was finally approved (not the questionable one, for whatever the hell reason, but I digress into my irritation) and I was targeting all the books by one author and set some fairly high bids (for me). One in particular I took way above the recommended bid.

Results: EoS is waaaayyy down the list, like between 15-17 flips of the carousel on a couple I checked, page 5 on another, and Ding! Ding! Ding! #1 on the book I jacked high.

I'll be exploring this further.

 

[Prince of Spires]

One explanation I've seen is that AMZ factors in both the bid and also the conversion rate. As in, 2 equal bids, but one has a higher conversion rate, then the one with the highest conversion rate ends up higher. Which factor is more important and how they each scale is hard to say without a lot more experimenting. And it might not even be constant. They're probably sliding scales.

It sort of makes sense, for 2 reasons. One is that AMZ looks at what makes them the most money. If you bid $2 per click, but never get any clicks then AMZ never makes any money off that placement. If it gets clicks but no sales, then AMZ only makes money from the clicks. However, if you have a $1 per click bid, and you convert every 4 clicks on a book which nets AMZ $2 per sale, then each click is worth $1.50. If those get twice the number of clicks as the $2 per click one, then they would make $3 for that ad compared to only $2 for the worse converting one. (if that still makes sense).

The other side I've heard argue is customer satisfaction. Not sure if this is really the case or not. But the idea goes that AMZ likes keeping customers happy. They do this by offering customers stuff they want to buy. It doesn't really matter what that stuff is, as long as they buy it. This means that accurately targeted products keep customers happier. Which means that those products will rank higher, even if they pay less per click.

People use this logic to "train the algorythm". They bid very high on great targets which gives them good conversions. The algorythm then learns that your book is a good match and that people are likely to buy it if they see it on that books page. Once you've got a decent number of sales this way, you can reduce your bid and still remain high in the list, even though you pay less than the ads around you.

 

[Demesnedenoir]

I'd agree to those possibilities. In the case of a tie, anyhow. Amazon certainly has a relevancy scale hidden in their algorithm and no doubt they use it. I was impressed with how quickly it took my new ad to account against a book and author I've never targeted before. Relevance is the overall killer of a great many ads.

This is something I started to play with a couple of years ago, but then just let slide. But back then I never hit a book with such a leap and a new target. Now that I have impressions (mostly) running steady and CTR getting more efficient, I figured it was time to get back to toying with this.